The present disclosure relates to an image processing device, an image processing method, and a program, and specifically relates to an image processing device, an image processing method, and a program which can suppress a false detection and a detection omission of a hindrance due to normalization of an image.
In the related art, a surveillance system which photographs a space which is a target using a surveillance camera, and detects an intruder from a photographed image, is generally used, in order to detect an intruder such as a person, an animal, or the like. In the surveillance system, it may be difficult to perform the surveillance, when there is a hindrance, for example, covering the surveillance camera with cloth, changing the direction of the surveillance camera, spraying something onto a lens of the surveillance camera, or the like.
A technique is proposed, in which a difference in a luminance value between the current image which is taken using the surveillance camera and the reference image which is stored in advance, is calculated, and it is determined that there is a hindrance when the difference value is equal to or greater than the fixed value (for example, refer to Japanese Unexamined Patent Application Publication No. 2008-77517).
FIG. 1 is a diagram which describes updating of an image which is photographed using the surveillance camera. As shown in FIG. 1, for example, an image which is obtained in every other frame is stored as the reference image, and is sequentially updated, when frames F1 to F6 are sequentially obtained. That is, in this example, frames F1, F3, and F5 are sequentially updated as the reference images.
When there is no hindrance, a difference in luminance value is small, since a background image temporally before is the same as a background image temporally after. In contrast, when there is a change in the direction of the surveillance camera due to a hindrance, the difference in luminance value is large, since the background image temporally before is different from the background image temporally after.
For this reason, when frame F3 is obtained, a difference in luminance value between frame F3 and frame F1 as the reference image which is a past image at the time, is calculated. Similarly to this, when frame F5 is obtained, a difference in luminance value between frame F5 and frame F3 which is an image in the past, is calculated. It is determined that there is a hindrance when a large difference is obtained.
Meanwhile, there is a case where the entire brightness (a luminance value) of the captured image using the surveillance camera, changes due to a change in illumination or weather, or an AE (Automatic Exposure) of the surveillance camera. In order to prevent a false detection due to the change in the entire brightness (the luminance value), a technique is disclosed, in which the images are compared to each other, after normalizing a histogram of the luminance value of the images (see, for example, Japanese Patent No. 3506934)